ltrn,d,

43, Uity,rchC-2

a b s t r a c t

required. These applications have a critical relationship betweensize of a mechanical system and the cost associated with manufac-turing and operation. Improvements could be made in the existingheat transfer systems by enhancing the performance of heat trans-fer uids resulting in lesser heat exchanger surface area, lower cap-ital costs, and higher energy efciencies. In this pursuit, numeroustechniques to enhance the thermal performance of heat transfer

nanotubes are suspended in conventional heat transfer uids,enhancements in thermal conductivity and convective heat trans-fer performance are observed [15]. The motivations behind thecurrent study are as follows. Firstly, there is limited informationabout the effect of preparation and processing conditions on thephysical properties and thermal performance of CNT aqueousnanouids. Secondly, limited experimental data is currently avail-able for CNT-based nanouids particularly in the area of viscosity.Lastly, only two studies have been reported to date on convectiveheat transfer of aqueous CNT nanouids [4,6]. Convective heat

* Corresponding author. Tel.: +1 979 458 1900; fax: +1 979 862 7969.

International Journal of Heat and Mass Transfer 52 (2009) 50905101

Contents lists availab

H

.eE-mail address: alvarado@entc.tamu.edu (J.L. Alvarado).1. Introduction

Managing high thermal loads has become very critical in therapid developed infrastructure, industrial, transportation, defense,space sectors. Several cooling technologies have been researchedrecently. Conventional heat transfer techniques that rely on uidslike water, ethylene glycol, and mineral oils continue to be populardue to its simple nature. Conventional heat transfer systems usedin applications like petrochemical, rening, and power generationare rather large and involve signicant amount of heat transfer u-ids. In certain cooling applications, small heat transfer systems are

uids have been investigated. One of the methods used is to addnanoparticles of highly thermally conductive materials like carbon,metal, metal oxides into heat transfer uids to improve the overallthermal conductivity. Nanoparticles could be either spherical orcylindrical. Carbon nanoparticles of cylindrical form are called car-bon nanotubes (CNT). One type of carbon nanotube is called multi-walled carbon nanotubes (MWCNT) because they have multipleconcentric tubes in a single conguration.

This study is concerned with nanouids prepared by dispersingMWCNT in water, and their potential use as heat transfer uids in ahost of applications. It has now been established that when carbona r t i c l e i n f o

Article history:Received 21 May 2008Received in revised form 17 April 2009Accepted 17 April 2009Available online 17 June 2009

Keywords:Convective heat transferNanouidsMulti-walledCarbon nanotubesThermal conductivityViscosityHeat transfer enhancementNon-Newtonian uidTEM0017-9310/$ - see front matter 2009 Elsevier Ltd. Adoi:10.1016/j.ijheatmasstransfer.2009.04.029Four samples of 1 wt% multi-walled carbon nanotube-based (MWCNT) aqueous nanouids prepared viaultrasonication were thermally characterized. Direct imaging was done using a newly developed wet-TEM technique to assess the dispersion state of carbon nanotubes (CNT) in suspension. The effect of dis-persing energy (ultrasonication) on viscosity, thermal conductivity, and the laminar convective heattransfer was studied. Results indicate that thermal conductivity and heat transfer enhancement increaseduntil an optimum ultrasonication time was reached, and decreased on further ultrasonication. The sus-pensions exhibited a shear thinning behavior, which followed the Power Law viscosity model. The max-imum enhancements in thermal conductivity and convective heat transfer were found to be 20% and 32%,respectively. The thermal conductivity enhancement increased considerably at temperatures greaterthan 24 C. The enhancement in convective heat transfer was found to increase with axial distance. Anumber of mechanisms related to boundary layer thickness, micro-convective effect, particle rearrange-ment, possible induced convective effects due to temperature and viscosity variations in the radial direc-tion, and the non-Newtonian nature of the samples are discussed.

2009 Elsevier Ltd. All rights reserved.e Laboratory for Computational Physics and Fluid Dynamics, Naval Research Laboratory, Code 6404, 4555 Overlook Ave. SW, Washington, DC 20375, USAAn experimental study on the effect of utransfer performance of multi-wall carbo

Paritosh Garg a, Jorge L. Alvarado b,*, Charles Marsh c

Kalyan Annamalai a

aDepartment of Mechanical Engineering, Texas A&M University, College Station, TX 778bDepartment of Engineering Technology and Industrial Distribution, Texas A&M UniverscU.S. Army, Engineer Research and Development Center, Construction Engineering ReseadDepartment of Nuclear, Plasma, and Radiological Engineering, 216 Talbot Laboratory, M

International Journal of

journal homepage: wwwll rights reserved.asonication on viscosity and heatnanotube-based aqueous nanouids

Thomas A. Carlson c, David A. Kessler e,

SA117 Thompson Hall, 3367 TAMU, College Station, TX 77843-3367, USALaboratory, 2902 Newmark, Drive, Champaign, IL 61822-1076, USA34, 104 South Wright Street, Urbana, IL 61801, USA

le at ScienceDirect

eat and Mass Transfer

l sevier .com/locate / i jhmt

b bulki inlet

at atransfer is an area which still needs to be completely explored andunderstood. In this study, an effort has been made to consider allthe above factors in a way that would move forward nanouid re-search to the next phase.

2. Background

2.1. Past research in CNT nanouids preparation

Nomenclature

K ow consistency index, mPa sn ow behavior indexN rotational speed, rpmk uid thermal conductivity, W/m Cx axial distance from the inlet of the test section, mh convective heat transfer coefcient, W/m2 Cq00s heat ux applied to the uid, W/m

2

Ts surface temperature, CTb uid bulk temperature, CP inner perimeter of the copper tube, m_m mass ow rate, kg/scp uid specic heat, kJ/kg CA inner surface area of the copper tube, m2

Di inside diameter of copper tube, mNu Nusselt numbere specic energy, J/gdp particle diameter, m or lmu uid velocity, m/sL heated length, cma particle radius, cmPr Prandtl numberRe Reynolds number, quDilPe Peclet number,

_cf d2paf

P. Garg et al. / International Journal of HeOne of the critical steps in preparing carbon nanouids is dis-persing carbon nanotubes in the base uid. Due to the high aspectratio of carbon nanotubes and strong Van der Waals forces be-tween carbon surfaces, dispersion of CNT in aqueous mediumcan be challenging. CNTs are hydrophobic in nature and thus can-not be dispersed in water under normal conditions. There are usu-ally two methods to disperse carbon nanotubes in base uids:mechanical and chemical [7]. Mechanical methods generally in-clude ultrasonication using a probe or a bath. Chemical methodsinclude using surfactants and CNT-functionalization using acids.The surfactant method changes the wetting or adhesion behaviorwhich helps in reducing their tendency to agglomerate. Chemicalfunctionalization generally involves treating CNTs with acids athigh temperature. This results in addition of polar groups like COOH or OH at defect sites on nanotube surface, thus makingCNTs more hydrophilic in nature. However, aggressive chemicalfunctionalization, can damage the nanotubes. Both mechanicaland chemical methods can reduce the aspect ratio distribution ofthe nanotubes. It has been reported that thermal conductivityenhancement in CNT nanouids decreases with reduction in aspectratio [1,8]; therefore, proper care still has to be taken during pro-cessing to minimize adverse effects. In this study, a combinationof a mechanical method through ultrasonication, and a chemicalmethod with surfactants were used.

Surfactants are used to disperse carbon nanotubes in severalcases. Some examples of previously used surfactants are sodiumdodecyl sulfate (SDS) [1], sodium dodecyl benzene sulfonate(SDBS) [2], hexadecyltrimethyl ammonium bromide (CTAB) [3]and Nanosperse AQ [3]. Through past studies, it was found thatSDBS failed at elevated temperatures [2]. Additionally, Gum Arabic(GA) was found to be a better surfactant than sodium dodecyl sul-fate (SDS) and cetyltrimethylammoniumchloride (CTAC) for dis-persing carbon nanotubes in DI water [9]. This has beenconrmed through testing where two samples of 1 wt% MWCNTaqueous suspensions were prepared using GA and SDS as surfac-tants. Suspensions prepared using GA were found to be visuallymore stable even after several weeks as compared to the ones pre-

o outletf uidGreek symbolss shear stress, N/m2_c shear rate, s1

s0

yield shear stress, N/m2

l uid viscosity, mPa sd hydrodynamic boundary layer thickness, mdt thermal boundary layer thickness, m_cf local mean shear rate experienced by uid, s1

af uid thermal diffusivity, m2/sq uid density, kg/m3

/ particle volume fractionx angular velocity of particle, rad/smf kinematic viscosity, cm2/s

Subscriptsp particles surface

nd Mass Transfer 52 (2009) 50905101 5091pared using SDS. Therefore, based on past and current researchactivities, GA was found to be the most suitable surfactant. How-ever, GA has a tendency to increase viscosity when added to DIwater. A highly viscous nanouid could also result in increasedpumping power for commercial applications. Therefore, it isimportant that the amount added is optimum.

2.2. Past research on viscosity of CNT nanouids

Viscosity of a heat transfer uid is important with respect tostudying its convective heat transfer and pumping power requiredfor practical applications. Optimization is required between heattransfer capability and the viscosity as it has a direct bearing onthe design of ow and heat transfer equipment. Experimental datafor the effective viscosity of aqueous nanouids is limited to cer-tain nanoparticles, such as Al2O3 [1013], CuO [13,14], TiO2 [10]and MWCNT [4]. Most of these studies have been directed towardsmetal oxide nanoparticles and there is only one work which stud-ied aqueous MWCNT extensively [4]. The parameters againstwhich viscosity has been studied until today include particle vol-ume concentration, temperature and shear rate. Empirical andaccurate analytical models for prediction of the viscosity of highaspect ratio nanouids are not currently available. Those studiesare mainly focused on spherical nanoparticles of metal oxides,and have their basis from the Einstein theory for viscosity [15]which takes into account spherical particles. In the case of CNTnanouids, such models cannot correlate the experimental datawell because the shape of CNTs does not satisfy the assumptionsin the Einstein model.

at aAqueous CNT nanouids have shown to exhibit shear thinningor pseudoplastic type of non-Newtonian behavior [4]. However,no study has been reported that correlates empirical data of CNTnanouids with theoretical non-Newtonian viscosity models. Thetheoretical models provide equations that correlate shear stressof a owing uid to shear rate. This helps in classifying the owbehavior of a new nanouid. The widely used models for non-New-tonian ow are Power Law, s K _cn and Herschel Bulkley,s s0 K _cn where s;K; _c; s0 and n are shear stress, ow consis-tency index, shear rate, yield shear stress and ow behavior index,respectively.

Additionally, no work has been done in studying the effect ofprocessing or ultrasonication time on the viscosity of MWCNTaqueous nanouids. Experimental work in this area could provideimpetus to theoretical model development for CNT nanouids. Inthis work, an effort has been made to study this effect by ttingthe experimental data in the form of a shear stress-shear ratemathematical model.

2.3. Past research on thermal conductivity of CNT nanouids

The heat transfer characteristic of a owing uid can be repre-sented by a Nusselt number, which takes into account the Prandtlnumber including thermal conductivity. Thus, a rst assessment ofthe heat transfer potential of a nanouid is to measure its thermalconductivity. To date, a lot of research data has been published inthis area for metal oxide nanouids but comparatively less for CNTnanouids [11,16,17]. One of the rst studies involving CNT nano-uids was by Choi et al. [5]. They measured the effective thermalconductivity of MWCNTs dispersed in synthetic poly(a-olen) oiland reported a thermal conductivity enhancement of 160% by1.0 vol % nanotubes in oil. Subsequently, data was published byXie et al. [18] where enhancements were reported for water, ethyl-ene glycol and decene as base uids. Assael et al. [1,3] data focusedon aqueous MWCNT nanouids with SDS, CTAB and NanosperseAQ as dispersants. However, both studies reported much lessenhancements as compared to those reported by Choi et al. [5].The maximum thermal conductivity enhancement observed byXie et al. [18] was only 20% for 1% nanotubes in decene by volume,and that observed by Assael et al. [1] was 38% for 0.6% CNTs inwater by volume. In 2004, Wen and Ding [2] published data usingSDBS as the dispersant. Their results were comparable to Xie et al.[18] and Assael et al. [1] and suggested that differences in interfa-cial resistances and thermal conductivities of carbon nanotubesused were the main reasons for the observed discrepancies withrespect to Choi et al. [5]. Additionally, the base uid used by Choiet al. [5] was poly-a-olen (a lower thermal conductivity thanwater), though percentage enhancement reported was high, theabsolute enhancement was not as high as expected. As SDBS wasalso found to fail at elevated temperatures, another set of datawas published using GA as dispersant by Ding et al. [4]. In theirstudy, a maximum enhancement of 79% was reported at 1 wt%MWCNT in water.

To date, most of the published data in MWCNT based nanouidsis focused on the thermal conductivity enhancement as a functionof particle volume concentration, base uid, and temperature. Ef-fects of particle size [1,3], dispersant (surfactant) [1,3] and acidity[4] have also been considered. Assael et al. [3] reported the effect ofparticle size indirectly by increasing the homogenization time byultrasonication, and concluded that when carbon nanotube sus-pensions are homogenized for long periods of time, their aspect ra-tio decreases, which subsequently decreases their thermalconductivity enhancements. Yang et al. [19] conducted similar

5092 P. Garg et al. / International Journal of Hestudies and reached analogous conclusions with oil dispersions.However, no other studies have been found to conrm these nd-ings. In the present study, thermal conductivity data with respectto ultrasonication time and temperature are presented.

Thermal conductivity enhancement in carbon nanotube disper-sions is still not completely understood. Several mechanisms andphenomena have been postulated that attempt to explain the ob-served enhancements. Studies have indicated that nanotubes con-duct current and heat ballistically or in fast diffusive manner [20].The ballistic conduction is associated with the large phonon mean-free path in nanotubes. Hence, carbon nanotubes should promotefaster heat diffusion in liquids. Furthermore, there is evidence thatan organized solid-like structure of a liquid at the interface is a po-tential governing mechanism in heat conduction from a solid wallto an adjacent liquid [21]. It has been postulated [5] that this orga-nized solid/liquid interface structure causes a favorable heat trans-port across the solidliquid interface. Additionally, Jang and Choi[22] postulated another theory using Brownian motion of nanopar-ticles as a potential mechanism for increased thermal conductivityof nanouids at elevated temperatures. It suggested that as tem-perature is increased, the viscosity of base uids is decreased andBrownian motion of nanoparticles is consequently increased. Ithas been postulated that convection-like effects are induced byBrownian motion which result in increased apparent thermal con-ductivities. However, Keblinski et al. [23] showed that Brownianmotion is unlikely to have direct role in the enhancement of ther-mal conductivity. Wen and Ding [2] suggested nanotube network-ing as one of the likely mechanisms that facilitates avenues forfaster diffusion and potential ballistic transport of energy carriers.From all those studies, it is difcult to agree upon a single mostimportant mechanism that solely contributes to enhanced thermalconductivity.

2.4. Past research on convective heat transfer of CNT nanouids

The practical utility of nanouids as heat transfer uids is bestdetermined by its convective heat transfer coefcient. Unlike pastresearch activities which have focused on thermal conductivity,the study of convective heat transfer of nanouids still needs tobe explored in more detail. Only few papers have been written inthis area, and most of them focused on metal oxide nanoparticles[10,11,16,17]. To date, only two papers have presented results fromMWCNT aqueous suspensions under constant heat ux and lami-nar ow conditions. Faulkner et al. [6] reported heat transferenhancement in a microchannel at very low Reynolds number(217) and particle volume concentrations between 1.14.4 vol%. Ding et al. [4] reported heat transfer enhancement at intermedi-ate Reynolds number (8001200) and low particle volume concen-tration (0.05 vol %). Although both of these papers reported heattransfer enhancement, however, the heat transfer enhancementtrends with respect to particle volume concentration in one paperwas found to contradict the other work. Therefore, it can be saidthat substantial amount of work is still required in this area. Boththese papers considered parametric effects of particle volume con-centration, Reynolds number and heat ux on the heat transferenhancement. However, ultrasonication time or particle sizereduction was not considered in those studies. As experimental re-search in this area is still new, theoretical models for heat transferenhancement are quite limited too.

3. Experimental design

3.1. Sample preparation

De-ionized (DI) water, Gum Arabic (GA) and multi-walled car-

nd Mass Transfer 52 (2009) 50905101bon nanotubes (MWCNT) were used to produce aqueous suspen-sions. The nanotubes were procured from Helix Material

Solutions Inc., USA. The nanotubes had a specied average outsidediameter of 1020 nm, length of 0.540 lm and purity of 95%, pro-duced by chemical vapor deposition (CVD) process. GA ne powderwas supplied by Biochemika. Four 500 g samples were preparedwith mass fraction of GA and MWCNT as 0.25% and 1%, respec-tively, however, with varying ultrasonication times including 20,40, 60, and 80 min. GA was dissolved in DI water using a magneticstirrer, followed by the addition of MWCNT to the solution. Theresulting composition was ultrasonicated for 5 min at 100% ampli-tude using a 130 W, 20 kHz ultrasonication probe (Sonics & Mate-rials, Inc., USA). As the probe sonicates within a limited conicvolume, to facilitate uniform dispersion, sonication was followedby 5 min of magnetic stirring. The ultrasonication and magneticstirring process were alternated every 5 min until the sample hadbeen sonicated for the desired amount of time. Based on processingtime, a certain amount of energy was transferred to each sample.

TEM technique developed by Dongxiang Liao and Jianguo Wenfrom the Frederick Seitz Materials Research Laboratory at the Uni-versity of Illinois at Urbana-Champaign was used [24]. A JEOL 2010LaB6 TEM was used with a beam acceleration voltage of 200 keV.The wet-cell was constructed by holding the uid between two sil-icon nitride membrane window TEM grids. The grids contained a200 lm thick frame and a 50 nm thick window, in which the sam-ple was placed. Thus, even a tiny drop of the sample could be used.The grids were then placed in a custom-built TEM sample holderthat included a number of o-rings intended to create a seal againstthe vacuum for the uid in the grid.

3.3. Viscosity measurement

The viscosity was measured using a low viscosity rotationaltype viscometer (LVDV-I Prime, Brookeld Engineering Laborato-

Sys

T

P. Garg et al. / International Journal of Heat and Mass Transfer 52 (2009) 50905101 5093This energy was divided by the mass of the sample (i.e., 500 g) toobtain the specic energy, e transferred to each sample. It was as-sumed that all the energy imparted was received by each uidsample.

Therefore, the samples are dened as:

(a) Sample A: 1 wt% MWCNT, 0.25 wt% GA, ultrasonicated for20 min, e = 57 J/g.

(b) Sample B: 1 wt% MWCNT, 0.25 wt% GA, ultrasonicated for40 min, e = 113 J/g.

(c) Sample C: 1 wt% MWCNT, 0.25 wt% GA, ultrasonicated for60 min, e = 188 J/g.

(d) Sample D: 1 wt% MWCNT, 0.25 wt% GA, ultrasonicated for80 min, e = 290 J/g.

The samples prepared by this technique were found to be stablefor over 1 month with no visible sedimentation or settling.

3.2. Wet-TEM imaging

One of the limitations of the conventional transmission electronmicroscopy (TEM) technique is that test samples have to be driedand exposed to vacuum before they can be imaged. This may in-duce structural changes in the sample as compared to the originaltest uid. Therefore, one can never be sure whether the dried testsample is representative of the original sample. To overcome thisproblem, a new type of TEM technique is now available knownas wet-TEM. Wet-TEM allows imaging of samples under wet orin-situ conditions without altering the original condition of the testuid. The new technique facilitates the imaging of the actual qual-ity of carbon nanotube dispersion within the base uid. A wet-cell

D.C. Power Supply

Data Acquisition

3,sT4,sTobT ,Fig. 1. Schematic for convective heat transries, Inc., USA). The model had a maximum torque rating of0.06737 mNm and a specied accuracy of 1%, which was veriedusing a Brookelds standard viscosity test uid. A combination ofcylindrical sample container and spindle called as UL Adapter wasused for taking measurements at low viscosity. The viscous dragexperienced by the spindle in UL Adapter was factory calibratedto display dynamic viscosity on a digital output screen. Measure-ments were taken at several shear rates at 15 and 30 C.

3.4. Thermal conductivity measurement

The thermal conductivity was measured using a KD 2 Pro ther-mal properties analyzer (Decagon devices, Inc., USA). The instru-ment had a probe of 60 mm length and a 1.3 mm diameter andincluded a heating element, a thermo-resistor and a microproces-sor to control and measure the conduction in the probe. The instru-ment had a specied accuracy of 5%. The samples were maintainedat several temperatures using a temperature-controlled chiller. Anumber of measurements were taken for each sample and a meanof only those measurements with R2 value (correlation coef-cient) greater than 0.9995 were considered. The instrument wasbased on the working principle of a transient hot wire method usedin past nanouid works [1,3,4,25].

3.5. Convective heat transfer measurement

The experimental set up used to measure convective heat trans-fer coefcient is shown schematically in Fig. 1. The set-up was cal-ibrated to give measurements within 5% accuracy. It consisted of acopper heat transfer section, data acquisition system, a DC powersupply, a syringe metering pump and a computer. A straight cop-

Syringe Pump

tem Computer

ibT ,1,sT2,s

Copper Tube Plastic Tube

fer measurement experimental set-up.

at aper tube of 914.4 mm length, 1.55 mm inner diameter and3.175 mm outer diameter was used as a test section. The wholesection was heated by an AWG 30 nichrome 80 wire (MWS wireindustries, USA) wound on the tube and connected to a 1500W,0300 V, 05 A DC power supply (Lambda, USA). Both ends of thecopper tube were connected to well-insulated plastic tubing toinsulate the heat transfer section and uid from axial heat conduc-tion, and to avoid heat losses. The experiments were run underconstant heat ux conditions using a current of 0.2 A. The test sec-tion was insulated to prevent loss of heat to the surroundings. Foursurface-mount thermocouples were mounted on the test section ataxial positions of 19 cm (Ts,1), 39.5 cm (Ts,2), 59 cm (Ts,3) and 79 cm(Ts,4) from the inlet of the section to measure wall temperatures.Additionally, two thermocouples were mounted on individual, un-heated, and thermally insulated, short copper tubes located beforeand after the heat transfer section to measure the uid bulk tem-perature at the inlet and outlet of the heat transfer section. Theuid was collected in a thermally insulated cup after passingthrough the heat transfer section where temperature was mea-sured under mixing cup conditions to validate the outlet bulk uidtemperature measured by the thermocouple. The pump used was aCole Palmer syringe metering pump. The ow rates used were 40,60 and 80 mL/min such that the ow conditions were always lam-inar. The corresponding Reynolds number for water at these owrates are approximately 600, 900 and 1200, respectively. All thethermocouples and the output from the DC power supply wereconnected to a data acquisition system (Agilent 34970A), whichwas connected to a computer. The set-up was calibrated both un-der isothermal and constant heat ux operating conditions, andcorrection factors were applied to all the measurements.

The convective heat transfer coefcient (h(x)) at an axial dis-tance x from inlet is dened as:

hx q00s

Tsx Tbx 1

where q00s ; Tsx and Tb(x) are heat ux applied to the uid, wall tem-perature at a distance x from the inlet, and uid bulk temperatureat a distance x from the inlet, respectively. From the energy bal-ance equation, the bulk temperature of the uid (Tb(x)) at an axialdistance, x can be found using:

Tbx Tb;i q00s :P_m:cp

x 2

where Tb;i; P; x; _m and cp are uid bulk temperature at the inlet,perimeter of the copper tube, axial distance from the inlet of thetest section, mass ow rate of the uid, and specic heat of the uid,respectively. The heat ux applied to the uid q00s can be foundusing:

q00s _m:cpTb;o Tb;i

A3

The convective heat transfer coefcient is also dened in the formof Nusselt number (Nu) as:

Nux hx Dik

4

where Di and k are the inside diameter of the copper tube and thethermal conductivity of the test uid, respectively.

4. Results and discussion

4.1. Imaging data

5094 P. Garg et al. / International Journal of HeFig. 2ad shows the pictures of samples AD at a scale of0.5 lm, under in-situ conditions, using wet-TEM technique [24].It can be seen that samples A and B exhibit a good three-dimen-sional network of carbon nanotubes. Samples C and D show shortercarbon nanotubes which can be attributed to the additional ultra-sonication time. From Fig. 2e and f (scale of 200 nm), it appearsthat the length of the nanotubes is relatively less in sample D thanin sample B. The images provide only snapshots of CNT nanouidsamples for different ultrasonication times. Though the imagesare not entirely representative of all nanouid samples, they givegood indication of the damage caused by excessive ultrasonication.The imaging data also provides some evidence that ow (viscosity)and thermal (i.e., thermal conductivity) properties should be af-fected by ultrasonication time as shown in the subsequentsections.

4.2. Viscosity data

A rotating drum viscometer was used to measure dynamic vis-cosity and shear rate. Fig. 3a and b shows the variation of viscositywith shear rate for each of the samples including water and0.25 wt% GA aqueous solution at 15 and 30 C, respectively. Itcan be clearly seen from the gure that that MWCNT aqueousnanouids displayed a non-Newtonian behavior especially at15 C even when taking variability into account as shown by theerror bars in Fig. 3a and b. A shear thinning or pseudoplastic behav-ior was observed resulting in a decrease in viscosity with an in-crease in shear rate up to 60 s1. In case of 0.25 wt% GA aqueoussolution, shear thinning was observed initially (up to 60 s1) buta slight shear thickening can be observed at 75 s1. A shear thin-ning effect can be explained by possible de-agglomeration of bun-dled nanotubes or realignment in the direction of the shearingforce, resulting in less viscous drag. A slight shear thickening canbe attributed to the unique uid properties Gum Arabic dispersionswhich have shown both shear thinning and shear thickeningbehavior at different shear rates in recent studies [26]. More workis required in this area to fully understand the role of Gum Arabicon nanouids.

Additionally, it is clear that the viscosity of the nanotube sus-pension rst increased from sample A to sample B, and thereafterdecreased with increase in ultrasonication time. According to Starret al. [27], a clustered nanoparticle suspension shows lower viscos-ity than a dispersed suspension. The increase in viscosity in a dis-persed sample is due to increased attractive surface interactions asa result of greater surface-to-volume ratio [27]. For a fully dis-persed sample, the total exposed nanoparticle surface is muchmore than in a clustered sample, resulting in greater viscosity indispersed samples. Furthermore, in this work it is suggested thatdue to less dispersing energy used in the preparation of sampleA, the nanotubes may not have received enough energy to over-come agglomerization and still remained in a clustered state. Onthe other hand, sample B was sonicated for more time, and re-ceived optimum energy to create a uniform dispersion which re-sulted in greater viscosity than sample A. After 40 min ofsonication, it was seen that the viscosity continuously decreasedwith further sonication. Another effect is the increased breakageof nanotubes with increase in ultrasonication time as observed inFig. 2. Excessive ultrasonication results in reduced nanotubeslength and aspect ratio. It has also been found by Yang et al. [19]that with increase in ultrasonication of carbon nanotube-in-oil dis-persions, viscosity decreases too. They explained this observationbased on the disruption of three-dimensional network of nano-tubes with a decrease in aspect ratios of nanotubes. This reasoningcan also be conrmed by examining all the wet-TEM pictures inFig. 2.

From the data, a non-Newtonian behavior is observed; however,

nd Mass Transfer 52 (2009) 50905101quantitative assessment requires a correlation between shearstress and shear rate by curve tting. Plots of shear stress vs. shearrates were made for each sample including 0.25 wt% Gum Arabic.

at aP. Garg et al. / International Journal of HeThe mathematical relation found for each sample was comparedwith the viscosity models discussed in Section 2.2, to nd out thecharacteristic ow behavior of the samples. It was found that allthe nanouid samples followed the Power Law s K: _cn viscositymodel. The R2 value (correlation coefcient) for each of the curveswas found to be more than 0.999, thus indicating a good correla-tion even when a slight shear thickening behavior was also ob-served at 75 s1. The unit of K is same as that of viscosity andtherefore, it is indicative of the magnitude of viscosity for a non-Newtonian uid. The values for ow consistency index, K and owbehavior index, n were found for each nanouid sample and0.25 wt% GA at 15 and 30 C, as shown in Fig. 4. From the gure,it is seen that ow consistency index increases with the ultrason-ication time from sample AB and thereafter, it decreases. This is inagreement with the viscosity trend observed in Fig. 3a and b. Addi-tionally, the corresponding values for 0.25 wt% GA aqueous solu-tion were found to be lower than the MWCNT nanouid samples,thus indicating that it is less viscous. From the Fig 4b, the owbehavior index (n) at 30 C remains almost constant with ultrason-ication time. Since a low value of n indicates a more pronouncednon-Newtonian behavior, from Fig 4b it can be said that there isa greater degree of non-Newtonian behavior at lower tempera-

Fig. 2. Wet-TEM images of aqueous suspensions of 1 wt% MWCNT, 0.25 wt% GA sonicatscale, (d) 80 min at 0.5 lm scale, (e) 40 min at 200 nm scale, and (f) 80 min at 200 nm snd Mass Transfer 52 (2009) 50905101 5095tures. However, at both temperatures it is observed that 0.25 wt%GA aqueous solution has a higher value of n than all nanouid sam-ples, thus indicating a lesser degree of non-Newtonian behavior.Therefore, the non-Newtonian behavior in MWCNT nanouid sam-ples is not just due to the presence of GA but due also by the inclu-sion of nanotubes.

4.3. Thermal conductivity data

Measurements were taken for all the nanouid samples includ-ing DI water at different temperature. All measurements for DIwater were found to be within 2% of the NIST values. The presenceof 0.25% Gum Arabic in water resulted in an insignicant change inthe thermal conductivity of water; therefore, water was used asbase uid for comparison purposes. From Fig. 5, the thermal con-ductivity of the nanouid samples rst increases slightly with tem-perature, and after 24 C it increases non-linearly withtemperature. One of the suggested reasons behind this phenome-non is the increased Brownian motion effect. Jang and Choi [22]suggested that as the temperature is increased, the viscosity ofthe nanouid decreases, which results in an increase in Brownianmotion of nanoparticles, which sets convection-like effects result-

ed for: (a) 20 min at 0.5 lm scale, (b) 40 min at 0.5 lm scale, (c) 60 min at 0.5 lmcale.

0.0

0.5

1.0

1.5

2.0

2.5

w c

onsi

sten

cy in

dex,

K ( m

Pa.se

c)

Temp = 15 C, MWCNT SamplesTemp = 30 C, MWCNT SamplesTemp = 15 C, 0.25 wt% GATemp = 30 C, 0.25 wt% GA

(a)

at and Mass Transfer 52 (2009) 509051011.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

10 20 30 40 50 60 70 80

Visc

osity

(mPa

.s)

Shear rate (sec-1)

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)0.25 wt % GADI water

0.9

1.0

1.1

1.2

1.3

scos

ity (m

Pa.s

)

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)0.25 wt % GADI water

(b)

(a)

5096 P. Garg et al. / International Journal of Heing in enhanced thermal conductivity. From the Fig. 5, a maximumincrease of 20% in thermal conductivity was obtained for sample Bat 35 C. The average density of MWCNT as provided by the vendorwas 2.1 g/cm3. Taking this value into account, a 1 wt% CNT suspen-sion would be approximately 0.47 by volume (vol %). The percent-age enhancement reported by Xie et al. [18] was 7% at 1 vol % ofMWCNT (aspect ratio 2000) and that by Assael et al. [1] was38% at 0.6 vol % MWCNT (aspect ratio 500). Therefore, it can besaid that both Xie el al. [18] and Assael et al. [1] used higher massfractions of CNTs than in this work. The enhancement values ob-tained in this work are much better than the value obtained byXie et al. [18], and slightly less than the value obtained by Assaelet al. [1]. However, Assael et al. [1] used a slightly higher volumefraction of MWCNT in his work which could have resulted in a bet-ter overall thermal conductivity value. However, the value re-ported by Ding et al. (i.e., 79% at 1 wt% MWCNT at 30 C) [4] ismuch higher than the values obtained in this work even whenthe concentration of CNTs used in both works is the same. The ex-act reason for this discrepancy is unclear. It could be suggested thatthe reason is related to the thermal and physical properties of theCNTs used in both works. Additionally, Ding et al. [4] did not spec-ify the CNT aspect ratio, and it could be different from the aspectratio (502000) used in this work. From Fig. 5, it can be inferredthat thermal conductivity enhancement is affected by ultrasonica-tion time, reaching an optimum for sample B. It is suggested thatthe reason for this phenomenon is associated with the aspect ratioof CNTs and the quality of three-dimension network establishedwithin each of the samples. It was found from wet-TEM imagingthat the aspect ratio of CNTs appears to have decreased with ultra-

0.7

0.8

10 20 30 40 50 60 70 80

Vi

Shear rate (sec-1)

Fig. 3. Variation of viscosity with shear rate at: (a) 15 C and (b) 30 C.20 40 60 80Flo

Ultrasonication Time (minutes)

0.94

0.96

0.98

1.00

beh

avio

r ind

ex, n

Temp = 15 C, MWCNT SamplesTemp = 30 C, MWCNT SamplesTemp = 15 C, 0.25 wt% GATemp = 30 C, 0.25 wt% GA

(b) sonication. This is evident in Fig. 2 where samples B and D werecompared at the 200 nm scale. Assael et al. [1] concluded that a de-crease in aspect ratio decreases thermal conductivity enhance-ment. Additionally, Wen and Ding [2] suggested carbon nanotubenetworking as one of the factors contributing towards enhancedthermal conductivity as it provides avenues for ballistic or fast heattransport. Sample A shows slightly lower thermal conductivityenhancement because there was not sufcient ultrasonication timeor energy to uniformly disperse nanotubes. On the other hand,sample B received sufcient energy for the creation of a uniformand effective three-dimensional network of CNTs. Hence, it canbe suggested that there was an optimum condition being estab-lished in sample B, where an optimal aspect ratio and a uniform

0.90

0.92

20 40 60 80

Flow

Ultrasonication Time (minutes)Fig. 4. Variation of: (a) ow consistency index with ultrasonication time and (b)ow behavior index with ultrasonication time.

0358

1013151820

10 15 20 25 30 35 40

% T

herm

al c

ondu

ctiv

ity

enha

ncem

ent

(k

/ kD

I wat

erX

100

)

Temperature (C)

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)

Fig. 5. Variation of percentage enhancement in thermal conductivity withtemperature.

three-dimensional network were reached. On further ultrasonicprocessing, this condition was disturbed which resulted in de-creased aspect ratio and a lower quality of three-dimensional net-work, resulting in a decrease in thermal conductivityenhancement.

4.4. Convective heat transfer data

For the heat transfer experiments, the heat ux was maintainedconstant at 0.6 W/cm2. The Reynolds numbers for DI water werefound to be approximately 600, 900 and 1200, respectively. Thepresence of 0.25% Gum Arabic in water resulted in an insignicantchange in the heat transfer coefcient of water in laminar owconditions; therefore, water was used as base uid for comparisonpurposes. However, in case of CNT nanouid as the viscosity of thesamples changed appreciably with temperature and shear rate(due to non-Newtonian behavior), the Reynolds number for thesesamples was found to vary within a range of 100. Figs. 6 and 7ashow the variation of convective heat transfer coefcient of allsamples and DI water with respect to non-dimensionalized axialdistance, x/Di for different Reynolds number. Figs. 6 and 7a alsoindicate the thermal entry length region and thermally fully devel-oped region. It can be seen that the laminar heat transfer coef-cient decreases with axial distance. This is expected due to theentry length phenomenon. For a pure Newtonian uid owingthrough a tube with a circular cross section, the ow is consideredto be hydrodynamically and thermally fully developed at x/

DiP 0.05.Re and x/DiP 0.05.Re.Pr, respectively. Once the owhas become fully developed, the heat transfer coefcient value sta-bilizes for pure uids. A similar trend is observed for CNT nanouidsamples to a certain extent. A clear enhancement in heat transfercoefcient was observed in the case of CNT nanouid samples ascompared to DI water.

The enhancement in convective heat transfer coefcient in-creases continuously with axial distance with the maximum in-crease found to be 32% for sample B at Re 600 100. Thisincrease is more than the maximum increase of 20% in thermalconductivity observed in sample B. These observations were foundto both agree and disagree with the previous work for nanouids.Ding et al. [4] reported similar trends where a more dramatic in-crease in heat transfer coefcient enhancement (i.e., 375% at0.5 wt% CNT) as compared to thermal conductivity enhancement(i.e., 37% at 0.5 wt% CNT) were observed. Similar trends were re-ported by Wen and Ding [16], and Xuan and Li [17]. Wen and Ding[16] reported a heat transfer coefcient enhancement of 47%, andthermal conductivity enhancement of 10% at 1.6 vol % Al2O3 spher-ical nanoparticles in DI water. Xuan and Li [17] reported a heattransfer coefcient enhancement of 60%, and thermal conductivityenhancement of 12.5% at 2 vol % using Cu spherical nanoparticles.In this work, in a similar way, the heat transfer coefcientenhancement has been found to be more than the thermal conduc-tivity enhancement. Ding et al. [4] suggested several possiblemechanisms for this observation. Using convective heat transferfundamentals, Ding et al. [4] stated that the heat transfer coef-cient, h can be approximately represented as k/dt where k and dt

(a)

2200

2400

sfe

r 2 -

K)

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)DI Water

Re ~ 600 +/- 100

Fig. 7. (a) Axial variation of heat transfer coefcient at Re 1200 100 and (b)microscopic picture of sample B (1 wt% MWCNT, 0.25 wt% GA, 40 min sonication) at100 lm.

P. Garg et al. / International Journal of Heat a(b)

1400

1600

1800

2000

0 200 400 600

Co

nve

ctiv

e he

at t

ran

coe

ffici

en

t, h(

x), (W

/m

x/Di

Thermal entry length

Thermally fully developed

region for DI water

1400

1600

1800

2000

2200

2400

0 200 400 600

Con

vect

ive

he

at t

ran

sfe

r co

effi

cie

nt,

h(x),

(W/m

2 .K)

x/Di

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)DI Water

Re ~ 900 +/- 100

Thermal entry length

for DI water

Thermally fully developed

regionfor DI WaterFig. 6. Axial variation of heat transfer coefcient at: (a) Re 600 100 and (b)Re 900 100.(a)

(b)

1300

1500

1700

1900

2100

2300

2500

0 200 400 600

Con

vect

ive

heat

tran

sfer

co

effic

ient

, h(x

), (W

/m2 .K

)

x/Di

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)DI Water

Re ~ 1200 +/-100

Thermal entry length region for DI

water

Thermally fully developed

region for DI water

nd Mass Transfer 52 (2009) 50905101 5097are the thermal conductivity of the test uid and thickness of thethermal boundary layer, respectively. With increase in k and de-crease in dt or d, the value of h should increase. A simultaneous de-

be a major mechanism behind the higher heat transfer enhance-ment as compared to thermal conductivity enhancement.

Ding et al. [4] observed that the enhancement of heat transfercoefcient reached a maximum for a certain value of x/Di. How-ever, in this work, we have found that the percentage enhancementincreases continuously with the axial distance as seen in Figs. 8and 9a. In Figs. 8 and 9a, the bulk temperature of the CNT nanouidincreases with axial distance, which results in a signicant increasein the thermal conductivity of the CNT nanouid as thermal con-ductivity increases with temperature (Fig. 5). As the heat transfercoefcient of a uid is directly proportional to thermal conductiv-ity in laminar ow (Eq. (4)), this in turn results in a slight increasein heat transfer coefcient. In addition to measuring the heat trans-fer coefcient, the local Nusselt number was calculated for eachsample to determine the net heat transfer enhancement in anon-dimensionalized way using Eq. (4). The corresponding thermalconductivity values for MWCNT nanouid samples were calculatedby linear interpolation from Fig. 5 after considering the bulk tem-perature from Figs. 8 and 9a. The Nusselt number values forRe 600 100 are shown in Fig. 9b. It was found that the experi-mental Nusselt number in the case of DI water matched well withthe theoretical fully-developed Nusselt number value of 4.36 for aconstant heat ux case. Further, from the Nusselt number calcula-tions, a clear heat transfer enhancement can be seen in Fig. 9b.

Figs. 10 and 11 show the variations for heat transfer coefcientenhancement with Reynolds number for all the nanouid samples.It can be seen the percentage heat transfer enhancement slightlydecreases with increase in Reynolds number. This is can be ex-plained by a decrease in bulk temperature of the nanouid at a par-ticular x/Di value with an increase in Reynolds number (Figs. 8 and9a). It was previously seen that the thermal conductivity enhance-

at acrease in d or dt could be suggested as one reason for the observedheat transfer enhancement. This could be explained by a possibleboundary layer thinning effect most likely caused by CNTs in theuid. However, in the fully developed laminar ow region dt andd are not expected to change appreciably, suggesting that otherenhancement mechanisms could be at play.

It has been found previously by Sohn and Chen [28] that for aliquid comprising of solid micro-scale particles, thermal conduc-tivity enhancements under shear conditions are greater thanthose observed under static conditions. This phenomenon wasattributed to micro-convective effects due to the presence ofan eddy-type convection mechanism. Signicant enhancementswere seen in their work for samples having Peclet number great-er than 300, where Peclet number, Pe is dened as

_cf d2paf; where

_cf ; dp and af are local mean shear rate experienced by uid, par-ticle diameter and thermal diffusivity of the uid, respectively.Additionally, Ahuja [29] found that signicant enhancementswere seen in thermal conductivity in tube ow when the Ahujanumber dened as / xamf

2 xa2af Ra

2 L2a 108 Doublet Collision

Frequency ratio where /, x, a, mf, R and L are particle volumefraction, angular velocity of particles, particle radius, kinematicviscosity, tube radius and heated length, respectively, has a valuenear to 0.02. The difference between both of these studies wasthat Sohn and Chen [28] study was based on couette owwhereas Ahuja [29] was based on poiseuille ow. However, bothstudies explained the enhancement due to the inertia of en-trained uid rotating with the particles. Based on a microscopicpicture of a drop of sample B (See Fig. 7b), it was found that thenanotubes in the present study microscopically exist as clusterswith a typical size between 10 and 20 lm. This value is muchless than the particle sizes of 2.9 and 0.3 mm used by Sohnand Chen [28], and 50 and 100 lm used by Ahuja [29]. Addition-ally, the particle volume fractions of suspensions used by themwere greater than those used in the current study (/ = 0.0047).The Peclet number as dened by Sohn and Chen [28] and theAhuja number [29] were calculated based on the typical clustersize (1020 lm), particle volume fraction, and ow conditions. Itwas found that the Peclet number (0.52.5) and Ahuja number(106107) were well below the values where signicantenhancement could be expected and explained by microconvec-tion. Therefore, a micro-convective effect could not be proposedas one of the major reasons for heat transfer enhancements inour study.

Also, CNTs in the nanouid could experience a re-arrangementeffect due to non-uniform shear rate in the radial direction. As seenin Figs. 3 and 5, the viscosity and thermal conductivity of CNTnanouids decreases with shear rate, and increases with tempera-ture, respectively. These lead to a non-uniform viscosity and ther-mal conductivity enhancement in the radial direction. Wen andDing [30] showed previously that such a change could result in ahigh Nusselt number and hence, a higher heat transfer coefcient.Thermal convection is improved if the viscosity near the wall of aowing uid in a tube is decreased with respect to the viscosity ofthe bulk uid [31]. In case of tube ow, the temperatures at thewall and the centerline are maximum and minimum, respectively.Due to these temperature variations, there are variations in viscos-ity in the radial direction, resulting in minimum viscosity at thewall and maximum viscosity at the centerline. This leads to con-vective effects in the radial direction, and hence, improved heattransfer coefcient [31]. Additionally, it has been found from pastwork [32] that a non-Newtonian uids have a higher Nusselt num-ber than Newtonian uids. Gingrich et al. [32] found that a uidwith a uid behavior index, n less than one (indicating shear thin-

5098 P. Garg et al. / International Journal of Hening behavior) exhibits higher heat transfer than one with n equalto unity. Since, CNT-based nanouids exhibit a shear thinningbehavior, it could be said that the non-Newtonian behavior could(a)

(b)

05

101520253035

0 100 200 300 400 500 600Perc

enta

ge E

nhan

cem

ent

(h

/ hw

ater

X 1

00)

x/Di

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)

Re ~ 900 +/-100

25.5 28 29.527Tb (oC)

30.524

05

10152025303540

0 200 400 600

Perc

enta

ge E

nhan

cem

ent

(h

/ hw

ater

X 10

0)

x/D i

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)

Re ~ 600 +/-100

Tb (oC)26 30 3228 3524

Fig. 8. Axial variation of percentage enhancement in heat transfer coefcient anduid bulk temperature at: (a) Re 600 100 and (b) Re 900 100.

nd Mass Transfer 52 (2009) 50905101ment has a considerable dependence on the bulk temperature ofthe nanouid. Therefore, due to a decrease in bulk temperatureof the nanouid, its thermal conductivity enhancement should also

(a)

(b)

0

5

10

15

20

25

30

0 100 200 300 400 500 600

Perc

enta

ge E

nhan

cem

ent

(h

/ hw

ater

X 10

0)

x/Di

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)

Re ~ 1200 +/-100

25 27 2826Tb (oC)

28.524

3.003.504.004.505.005.506.00

0 100 200 300 400 500 600

Expe

rimen

tal N

usse

lt nu

mbe

r, N

u(x)

x/Di

Sample A (20 min)Sample B (40 min)Sample C (60 min)Sample D (80 min)DI Water

Re~ 600 +/-100

Thermal entry length regionfor DI water

Thermally fully developed region

for DI water

4.36

P. Garg et al. / International Journal of Heat adecrease. This results in slight a decrease in heat transfer enhance-

Fig. 9. (a) Axial variation of percentage enhancement in heat transfer coefcientand uid bulk temperature at Re 1200 100, (b) variation of experimentalNusselt number with axial distance at Re 600 100.ment. However, as this enhancement reduction was found to besmall, further investigation is required.

From Figs. 8 and 9a, it can be inferred that ultrasonication timeindirectly affects heat transfer enhancement. The maximum

(b)

02468

101214161820

500 700 900 1100 1300

Per

cent

age

Enh

ance

men

t (

h / h

wat

er X

100

)

Re

x/Di = 123x/Di = 255x/Di = 381x/Di = 510

Sample A (20 min)

0

5

10

15

20

25

30

35

40

500 700 900 1100 1300

Per

cent

age

Enh

ance

men

t(

h / h

wat

erX

100

)

Re

x/Di = 123x/Di = 255x/Di = 381x/Di = 510

Sample B (40 min)

(a)

Fig. 10. Variation of percentage enhancement in heat transfer coefcient withReynolds number: (a) sample A and (b) sample B.enhancement was observed in sample B. This phenomenon is ob-served at all the three Reynolds number used in the study. A sim-ilar trend was seen in the thermal conductivity data. The reason forthe phenomenon could be again associated with the aspect ratio ofCNTs and the quality of three-dimension network establishedwithin each of the samples, as described in Section 4.3.

5. Conclusions and recommendations for future work

5.1. Conclusions

Through this work, it has been conrmed once again that CNTnanouids have potential as heat transfer uids. The aqueous sus-

05

10152025303540

500 700 900 1100 1300

Perc

enta

ge E

nhan

cem

ent

(h

/ hw

ater

X 10

0)

Re

x/Di = 123x/Di = 255x/Di = 381x/Di = 510

Sample C (60 min)

05

10152025303540

500 700 900 1100 1300

Perc

enta

ge E

nhan

cem

ent

(h

/ hw

ater

X 10

0)

Re

x/Di = 123x/Di = 255x/Di = 381x/Di = 510

Sample D (80 min)

(a)

(b)

Fig. 11. Variation of percentage enhancement in heat transfer coefcient withReynolds number: (a) sample C and (b) sample D.

nd Mass Transfer 52 (2009) 50905101 5099pensions of multi-walled carbon nanotubes prepared by usingGum Arabic as surfactant were found to be stable for months.However, the preparation method of the nanouids is an importantstep affecting the overall heat transfer performance. It was foundthat at a given CNT composition, there are certain optimum pro-cessing conditions (ultrasonication time in this case) which givethe maximum enhancement in heat transfer performance. Theowing conclusions are presented:

Ultrasonication has twofold effect on the CNT nanouids. Belowthe optimum processing time, the ultrasonication aids in form-ing better dispersions, however, once the optimum time hasbeen reached further ultrasonication results in an increasedbreakage rate of the nanotubes, and hence reduces the aspectratio of CNTs. In this work, the optimum ultrasonciation timewas found to be 40 min at 1 wt% MWCNT concentration, usinga 130W, 20 kHz ultrasonicator.

Viscosity of the nanouids increases with sonication time until amaximum value is reached and decreases thereafter. The initialincrease is associated with declustering of CNT bundles, result-ing in formation of a better dispersion. The later decrease in vis-cosity can be explained by increased breakage rate of CNTs,resulting in shorter nanotubes, and hence, inferior networkingof CNTs in dispersion.

CNT nanouids exhibit a shear-thinning (pseudoplastic) non-Newtonian behavior and follow the Power Law viscosity model.The ow consistency index was found to be higher than that of0.25 wt% Gum Arabic aqueous solution. Also, the ow behavior

at aindex was found to be lower than that of 0.25 wt% Gum Arabicaqueous solution. This shows that the non-Newtonian characterin CNT nanouids could not be solely attributed to the presenceof Gum Arabic.

The maximum percentage enhancement in thermal conductivitywas 20%, and was observed in Sample B which was sonicated for40 min.

The percentage enhancement in thermal conductivity of CNTnanouids increases considerably after 24 C.

The maximum thermal conductivity enhancement was obtainedfor an ultrasonication time of 40 min, and was found to decreasewith further sonication. The initial increase was explained byformation of better three-dimensional network in the CNT sam-ples, and the later decrease was explained by a decrease inaspect ratio of CNTs.

The maximum percentage enhancement in heat transfer coef-cient was 32% at Re 600 100 and was observed in Sample B.

The percentage enhancement in heat transfer coefcient at aparticular axial distance was found to slightly decrease withan increase in Reynolds number. This effect was attributed tothe decrease in thermal conductivity enhancement due to lowerbulk temperature of the uid which decreases with Reynoldsnumber. However, as this decrease was slight, further investiga-tion is still required.

The percentage enhancement in heat transfer coefcient wasfound to continuously increase with axial distance. The reasonbehind the phenomenon is explained by the contribution froma considerable increase in thermal conductivity with an increasein bulk temperature with axial distance.

The maximum percentage enhancement in heat transfer coef-cient (i.e., 32%) was found to be more than the maximum per-centage enhancement in thermal conductivity (i.e., 20%). Manypossible mechanisms were studied and presented for this phe-nomenon including the contribution of boundary layer thick-ness, micro-convective effect, particle rearrangement, possibleinduced convective effects due to temperature and viscosityvariations in radial direction, and non-Newtonian nature of thesamples.

5.2. Recommendations for future work

Though from experimental work it is known that nanouids ex-hibit enhanced heat transfer performance, the contribution of eachproposed physical mechanism still is unknown. Computer simula-tions that take into account the non-Newtonian characteristic ofthe nanouids could give a better understanding of the observedbehavior. Other parameters like base uid type (ethylene glycol,mineral oil, refrigerants), aspect ratio, ow conditions (laminar,transition, turbulent) could provide further understanding of theheat transfer behavior of CNT-based heat transfer uids.

The experimental data reported previously for nanouids haslarge variations. This is due to lack of standardized proceduresfor preparing nanouids. An extensive work in this area could helpin future research activities as well as supporting commercializa-tion efforts in the area of nanouids for heat transfer applications.

Acknowledgements

The authors express their deepest thanks to Alex Hays and RyanFranks of the U.S. Army Corps of Engineers Construction Engineer-ing Research Laboratory; Guillermo Soriano and Landon Sommerfrom Texas A&M University; Professor B. Jones of the Nuclear, Plas-

5100 P. Garg et al. / International Journal of Hema, and Radiological Engineering Department, and Jianguo Wen,Dongxiang Liao of the Frederick Seitz Materials Research Labora-tory at the University of Illinois at Urbana-Champaign for theirsupport. The project was supported by the National Science Foun-dation, SBIR/STTR program, and the U.S. Army Corps of Engineers.

References

[1] M.J. Assael, C.F. Chen, I. Metaxa, W.A. Wakeham, Thermal conductivity ofsuspensions of carbon nanotubes in water, Int. J. Thermophys. 25 (4) (2004)971985.

[2] D. Wen, Y. Ding, Effective thermal conductivity of aqueous suspensions ofcarbon nanotubes (carbon nanotube nanouids), J. Thermophys. Heat Transfer18 (4) (2004) 481485.

[3] M.J. Assael, I. Metaxa, J. Arvanitidis, D. Christolos, C. Lioustas, Thermalconductivity enhancement in aqueous suspensions of carbon multi-walled anddouble-walled nanotubes in the presence of two different dispersants, Int. J.Thermophys. 26 (3) (2005) 647664.

[4] Y. Ding, H. Alias, D. Wen, R.A. Williams, Heat transfer of aqueous suspensionsof carbon nanotubes (CNT nanouids), Int. J. Heat Mass Transfer 49 (2006)240250.

[5] S.U.S. Choi, Z.G. Zhang, W. Yu, F.E. Lockwood, E.A. Grulke, Anomalous thermalconductivity enhancement in nanotube suspensions, Appl. Phys. Lett. 79 (14)(2001) 22522254.

[6] D.J. Faulkner, D.R. Rector, J. Davidson, R. Shekarriz, Enhanced heat transferthrough the use of nanouids in forced convection, in: Proceedings of IMECE,Springer, Berlin, 2004.

[7] J. Hilding, E.A. Grulke, Z.G. Zhang, F. Lockwood, Dispersion of carbon nanotubesin liquids, J. Dispersion Sci. Technol. 24 (1) (2003) 141.

[8] R.L. Hamilton, O.K. Crosser, Thermal conductivity of heterogeneous two-component systems, IEC Fundam. 1 (3) (1962) 187191.

[9] R. Bandyopadhyaya, E. Nativ-Roth, O. Regev, R.Y. Rozen, Stabilization ofindividual carbon nanotubes in aqueous solutions, Nano Lett. 2 (1) (2002) 2528.

[10] B.C. Pak, Y.L. Cho, Hydrodynamic and heat transfer study of dispersed uidswith submicron metallic oxide particles, Exp. Heat Transfer 11 (2) (1998) 151170.

[11] S.K. Das, N. Putra, W. Roetzel, Pool boiling characteristics of nano-uids, Int. J.Heat Mass Transfer 46 (5) (2003) 851862.

[12] C. Li, M. Akinc, J. Wiench, M. Pruski, C.H. Schilling, Relationship between watermobility and viscosity of nanometric alumina suspensions, J. Am. Ceram. Soc.88 (10) (2005) 27622768.

[13] S.Z. Heris, S.G. Etemad, M.N. Esfahany, Experimental investigation of oxidenanouids laminar ow convective heat transfer, Int. Commun. Heat MassTransfer 33 (4) (2006) 529535.

[14] D.P. Kulkarni, D.K. Das, G.A. Chukwu, Temperature dependent rheologicalproperty of copper oxide nanoparticles suspension (nanouid), J. Nanosci.Nanotechnol. 6 (4) (2006) 11501154.

[15] A. Einstein, Eine neue Bestimmung der Molekuldimension, Annalen der Physik19 (1906) 289306.

[16] D.S. Wen, Y.L. Ding, Experimental investigation into convective heat transfer ofnanouids at entrance area under laminar ow region, Int. J. Heat MassTransfer 47 (24) (2004) 51815188.

[17] Y.M. Xuan, Q. Li, Investigation on convective heat transfer and ow features ofnanouids, ASME J. Heat Transfer 125 (2003) 151155.

[18] H. Xie, H. Lee, W. Youn, M. Choi, Nanouids containing multiwalled carbonnanotubes and their enhanced thermal conductivities, J. Appl. Phys. 94 (8)(2003) 49674971.

[19] Y. Yang, E.A. Grulke, Z.G. Zhang, G. Wu, Thermal and rheological properties ofcarbon nanotube-in-oil dispersions, J. Appl. Phys. 99 (2006) 1114307-11114307-8.

[20] S. Berber, Y.K. Kwon, D. Tomanek, Unusually high thermal conductivity ofcarbon nanotubes, Phys. Rev. Lett. 84 (20) (2000) 46134616.

[21] T. Ohara, D. Suzuki, Intermolecular energy transfer at a solidliquid interface,Nanoscale Microscale Thermophys. Eng. 4 (3) (2000) 189196.

[22] S.P. Jang, S.U.S. Choi, Role of brownian motion in the enhanced thermalconductivity of nanouids, Appl. Phys. Lett. 84 (21) (2004) 43164318.

[23] P. Keblinski, S.R. Phillpot, S.U.S. Choi, J.A. Eastman, Mechanisms of heat ow insuspensions of nano-sized particles (nanouids), Int. J. Heat Mass Transfer 45(4) (2002) 855863.

[24] R. Franks, S. Moreeld, W. Jianguo, D. Liao, J. Alvarado, M. Strano, C. Marsh, Astudy of nanomaterial dispersion in solution by wet-cell transmission electronmicroscopy, J. Nanosci. Nanotechnol. 8 (14) (2008).

[25] A. Alloush, W.B. Gosney, W.A. Wakeham, A transient hot wire instrument forthermal conductivity measurements in electrically conductivity instrumentsin elevated temperatures, Int. J. Thermophys. 3 (3) (1982) 225235.

[26] C. Sanchez, D. Renard, P. Robert, C. Schmitt, J. Lefebvre, Structure andrheological properties of acacia gum dispersions, Food Hydrocolloids 16(2002) 257267.

[27] F.W. Starr, J.F. Douglas, S.C. Glotzer, Origin of particle clustering in a simulatedpolymer nanocomposite and its impact on rheology, J. Chem. Phys. 119 (3)(2003) 17771788.

[28] C.W. Sohn, M.M. Chen, Microconvective thermal conductivity in disperse twophase mixture as observed in a low velocity couette ow experiment, J. HeatTransfer Trans. ASME 103 (1981) 4751.

nd Mass Transfer 52 (2009) 50905101[29] A.S. Ahuja, Augmentation of heat transfer in laminar ow of polystyrenesuspensions. II. Analysis of the data, J. Appl. Phys. 46 (8) (1975) 34173425.

[30] D.S. Wen, Y.L. Ding, Effect on heat transfer of particle migration in suspensionsof nanoparticles owing through minichannels, Microuidics Nanouidics 1(2) (2005) 183189.

[31] W. Kamil, Heat transfer in temperature-dependent non-newtonian ow,Chem. Eng. Process. 43 (2004) 12231230.

[32] W.K. Gingrich, Y.I. Cho, W. Shyy, Effect of shear thinning on laminar heattransfer behavior in a rectangular duct, Int. J. Heat Mass Transfer 35 (11)(1992) 28232836.

P. Garg et al. / International Journal of Heat and Mass Transfer 52 (2009) 50905101 5101

An experimental study on the effect of ultrasonication on viscosity and heat transfer performance of multi-wall carbon nanotube-based aqueous nanofluidsIntroductionBackgroundPast research in CNT nanofluids preparationPast research on viscosity of CNT nanofluidsPast research on thermal conductivity of CNT nanofluidsPast research on convective heat transfer of CNT nanofluids

Experimental designSample preparationWet-TEM imagingViscosity measurementThermal conductivity measurementConvective heat transfer measurement

Results and discussionImaging dataViscosity dataThermal conductivity dataConvective heat transfer data

Conclusions and recommendations for future workConclusionsRecommendations for future work

AcknowledgementsReferences